Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2022-05-18 Cooperative journals: 《计算机应用研究》
Abstract： Synthetic minority oversampling technology (SMOTE) is one of the effective methods to solve the class-imbalanced problem. However, the linear interpolation mechanism of SMOTE restricts the synthesized samples to the connecting line of the original samples, resulting in a lack of diversity for new samples, and may generate noisy samples when this line passes through the majority class region. In response to the above issues, this paper proposed a generation mechanism for minority samples with hypercuboid constraints. This mechanism constructed a hypercuboid as the generation region of new samples instead of linear interpolation, thereby increasing the variability between the synthesized samples and the original samples. Then, it detected whether there were majority samples in the hypercuboid to determine whether to adjust the hypercuboid, which aimed at preventing the new samples into the region of the majority class. This paper integrated the proposed mechanism into three oversampling methods, i. e. , SMOTE, Borderline-SMOTE, and ADASYN, by using it to replace linear interpolation, and then experimentally evaluated the integrated methods on 11 benchmark datasets from KEEL. The results showed that compared to the original methods, the integrated methods could help the classifier to obtain higher F1 and comparable G-mean. It verifies that the hypercuboid generation mechanism can significantly improve the classifier’s ability to recognize minority samples, and meanwhile the majority samples are also taken into account.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-01-03 Cooperative journals: 《计算机应用研究》
Abstract： Aiming at the sparseness and low prediction accuracy of traditional collaborative filtering algorithms, this paper proposed a hybrid recommendation algorithm based on restricted Boltzmann machine and weighted Slope One. Firstly, it use the preliminary filling of the scoring matrix by the restricted Boltzmann machine to alleviate the sparseness problem of the data. Then, it introduced the project attribute information through a hybrid project similarity calculation method. Finally, it adopted the second prediction by the weighted Slope One algorithm to improve the recommended effect. Experiments on the MovieLens100K dataset show that the combination of the two algorithms increases the accuracy of the recommendation.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-12-13 Cooperative journals: 《计算机应用研究》
Abstract： In order to solve the shortcomings of current Hadoop cluster inherent scheduling distributed methods in heterogeneous resource environments, this paper proposed an adaptive scheduling algorithm NCAS (Node Capacity Adaptive Scheduling) based on the node capability. Firstly, NCAS algorithm calculated the scheduling factor based on node performance and task characteristics; Secondly, the scheduling factor determined the amount of data and the number of task slot that each node should be assigned; Finally, NCAS algorithm dispatched data and tasks more into fast nodes and less into slow nodes. Experimental results show that, compared with the traditional scheduling algorithm, NCAS algorithm can greatly reduce the number of speculative tasks, significantly reduce the job completion time. It also can improve the task execution efficiency .
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-11-29 Cooperative journals: 《计算机应用研究》
Abstract： The location-based hierarchical access scheme based on attribute encryption allows users to flexibly set their own location access information according to their own situation. Not only solving the problem of location sharing in social networks, but also improving the algorithm to improve the decryption efficiency. However, during the operation of the system, there is a possibility that the user has corrected his own attribute information or the private key may be leaked during the operation, supporting the withdrawal is very necessary for system security. Based on this, a location hierarchical attribute encryption scheme supporting undo is proposed, which outsources part of the decryption operation to the decryption server and combines the method of two-factor identity authentication. This solution reduces the user's computational cost and improves the security of the algorithm.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-10-11 Cooperative journals: 《计算机应用研究》
Abstract： The BiLSTM neural network model has long training time and can not fully learn text context informations. In order to solve the problems, this work proposes a text emotion classification model based on BiGRU-Attention neural network. Firstly, the bidirectional Gated Recurrent Unity (BiGRU) neural network layer was used to extract the features of the deep text information; secondly, the attention mechanism (Attention) layer was used to allocate the corresponding weight of the extracted text deep information. Finally, the text feature messages of different weights are put into the softmax function layer to carry out the text sentiment classifications. The experimental results show that the accuracy of the proposed neural network model is 90.54% on the IMDB data set, the loss rate is 0.2430 and the time cost is 1100s and the validity of the BiGRU-Attention model is verified.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-08-13 Cooperative journals: 《计算机应用研究》
Abstract： In order to solve the problem that k-means clustering algorithm is dependent on the initial value and easily falls into the local optimum, this paper proposed a K-means clustering algorithm based on improved flower pollination. Firstly, the algorithm used the chaotic map sequence as the initial position of the flower population to ensure the diversity and determinacy of the flower population in the search space; Then, it introduced a tabu search algorithm in the late stage of flower pollination to avoid falling into the local optimal solution; Finally, used the improved flower pollination algorithm to optimize the initial value of the k-means algorithm. Experimental results on five clustering datasets show that the improved algorithm improves the average clustering accuracy by 12.2% compared with the flower pollination clustering algorithm, which proves that the proposed algorithm has better clustering performance for low-dimensional datasets.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-08-13 Cooperative journals: 《计算机应用研究》
Abstract： Collaborative filtering recommendation, which combines metadata, is a hot topic in recommender systems, which can solve data sparsity and cold-start problems to some extent. However, most of the existing modeling methods of combines metadata are based on the same user / item attribute weights, so that the key relationships between users and items are not significant, and it is difficult to obtain better recommendation performance. To solve the above problems, this paper proposes a method of deep joint learning recommendation based on metadata and attention mechanism. It uses double deep network learning, one of the networks implements matrix nonlinear decomposition based on implicit feedback data to learn user/project personalization relationships. , and another network automatically capture affect the user/item key attributes to recommend by using Attention mechanism, through the user preference relation with weights of different attributes modeling highlights the extended model. Experimental results show that the proposed recommendation algorithm has better recommendation performance on two public datasets of MovieLens 100K and MovieLens 1M.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》
Abstract： The attribute encryption scheme based on file hierarchy is efficient and low in the cloud storage environment, but the access structure itself contains sensitive information, there is the risk of user information leakage and easy to be stolen. Aiming at this problem, this paper proposed a file-level attribute encryption scheme with hidden access structure. The scheme improved the security of the encryption algorithm without affecting the encryption and decryption efficiency, and adopted a two-factor authentication mechanism to achieve a more secure and efficient access control. The results of the study are based on the deterministic bilinear Diffie-Hellman hypothesis, which is proved to be safe under the standard model.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》
Abstract： This paper studied the information packets multicast in communication networks, it proposed the re-transmission strategy based on cooperative relay in communication networks(CRRS) . When the packet fails to transmit directly, the source node, coordinated with the relaying nodes can encode and then re-transmit lost packets. During re-transmission, those lost packets accessible to encoding will assemble by way of random access according to the results derived from the feedback system. Subsequently, there will provide more spacial diversity gain for information restoration by obsoleting the nodes previously failed to transmit so as to reduce the frequency of the re-transmission. Finally, through diverse information channels, this strategy can simulate with non-coordinated NCARQ and traditional ARQ in Monte Carlo. The simulation results have shown that, on the condition that multiple relaying channels provide better conditions than source-destination channels, the use of coordination-based network encoding for the purpose of re-transmission would not only effectively promote network throughput capability, but also largely reduce the possibilities of poor performances due to coherence issues by way of coordinated spacial diversity.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》
Abstract： Making use of semantic information to achieve the high-level analysis and understanding is a hot issue currently. To address the problem of automatic recognition in the presence of significant geometric and topological variations, this paper proposed a 3D shape function recognition method by adopting the contextual relationship of shape parts. Firstly, it decomposed 3D shapes into the shape part sets with independent semantics and the technique of approximate convexity analysis could be employed. Then, it computed the contextual relationship of shape parts and on this basis, Support Vector Machines could be adopted to achieve the task of automatic recognition between shape parts. Experimental results show that the proposed method achieves higher the matching accuracy values and lower classification error rates, compared to the existing methods.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-18 Cooperative journals: 《计算机应用研究》
Abstract： The existing subspace clustering algorithms cannot balance the density of the data in the same subspace and the sparsity of the data between different subspaces and most algorithms cannot solve the overlap of data. To solve the above problems, this paper proposed a novel algorithm of overlapping subspace clustering algorithm under sparse condition (OSCSC) . The algorithm used the mixed norm representation method of L1 norm and Frobenius norm to establish the subspace representation model, and the weighted L1 norm regular term could improve the sparsity of different subspaces and the density of the same subspace. Then, the algorithm performed rechecks on the partitioned subspaces by using an overlapping probability model subject to exponential family distribution to determine whether exist overlapping in different subspaces, which could further improve the accuracy of clustering. The results of the experiment on both artificial datasets and real-world datasets show that the algorithm has better clustering performance by being compared to other contrast algorithms.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-24 Cooperative journals: 《计算机应用研究》
Abstract： Aiming at the low accuracy of direct inverse neural network model, the poor generalization ability of BP neural network, and the increased design time due to the fact that antenna parameters need to be optimized continuously using HFSS simulation software individually, this paper proposed a new method , which combined HFSS with sparse regularization inverse neural network to design the dual band-notched ultra-wide band antenna. This method added L1/2 norm and L2 norm in the performance function of the inverse neural network. The L1/2 norm introduced a new weight coefficient, expanded the input sample vector, made the network to obtain the sparsity solution more easily, and the inverse model got higher accuracy. Meanwhile, the L2 norm avoided the over-fitting phenomenon effectively and made the network generalization ability stronger. It applied to the design of dual band-notched ultra-wide band antenna, using of arc grooves on radiating patches generated notch characteristics, and according to the antenna target voltage standing wave ratio(VSWR) solved inversely the corresponding slot size. Simulation results show that the relative error of slot angle which corresponding to VSWR of the antenna reduced by 69.3%, and the relative error of slot radius lessened by 88.7%, and the network running time decreased by 15.9% compared with BP neural network method. The final designed antenna bandwidth is 2.4~11 GHz, achieves the good notch characteristics in 3.31~3.8 GHz and 4.98~6.05 GHz, and shortens the entire antenna design cycle.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-24 Cooperative journals: 《计算机应用研究》
Abstract： Aiming at typical attack types of industrial control networks, this paper proposed a method of predicting communication anomalies in industrial networks using deep learning. First, the principal component analysis of the raw data reduction and eliminated the correlation between the original data set. Secondly, build artificial neural networks and to optimize the input weights and threshold limits the use of machine learning. The fish swarm algorithm was improved by the idea of particle inertia mass calculation in the gravitational search algorithm. The test experiment results show that the accuracy of anomaly detection is improved, and the detection times are effectively shortened. And realizes the purpose of making use of the depth learning to predict the abnormal behavior of communication in industrial networks.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-24 Cooperative journals: 《计算机应用研究》
Abstract： In the mass music, how to analyze the user's needs according to the user's history listening record to implement song recommendation is one of the challenging topics in the music recommendation field. The existing music recommendation method simply uses all the music the user has heard as the context of the music recommendation, which results in the same weight distribution of contexts learned by different types of music, which seriously affects the accuracy of the music recommendation. In response to this problem, this paper proposed a music recommendation method based on attention mechanism, which dynamically allocated different attentions to different users' historical listening music, that was learns different contextual weights so as to make the recommendation result more in line with the user's actual preference . And through the test on public music dataset named Million Song Dateset, the recommended accuracy of the method proposed in this article has greatly improved.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-19 Cooperative journals: 《计算机应用研究》
Abstract： Although people have done a lot of research on cloud simulation, the existing cloud simulation algorithm can hardly obtain the cloud image with strong realism while ensuring the real-time performance. According to this problem, this paper proposes a cumulus-based simulation method for particle systems that combine Qsplat with IFS. First, build a tree-structured particle system based on the detail-level tree structure of Qsplat for speeding up particle system search speed. Then, for these particles which closer to the viewpoint , used the iterative function system algorithm to increase the level of enclosing the ball tree to enrich the local detail of the cloud to establish the shape of the cloud. Finally, Finally, rendered the cloud by using two-dimensional texture mapping of spherical entities, which realizes the fast real-time simulation of dynamic cumulus. The experimental results show that the method can generate real-time three-dimensional dynamic cumulus images quickly.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-12 Cooperative journals: 《计算机应用研究》
Abstract： Aiming at the sparse user rating data, domestic and foreign scholars have made many improvements on collaborative filtering algorithm, which were summarized as filling user rating data, improving similarity, fusing content to recommend and so on. These single methods can’t solve the problem of data sparseness. In order to solve this problem, this paper proposed a collaborative filtering algorithm which combines the filling data and improving similarity. Firstly, it used the improved filling method which increases the item’s attribute information to fill the user rating data, and then recommended using new similarity method, produced the recommended results, iterated m times. Finally it recommended items according to the average score of scores got in m iterations. The experiment shows that the proposed algorithm has a better recommendation effect than single methods in the case of sparse user rating data.
Hits
Hits
Downloads
Comment